An Augmented Model with Inferred Blood Features for the Self-diagnosis of Metabolic Syndrome.
Tianshu Zhou,Ying Zhang,Chengkai Wu,Chao Shen,Jingsong Li,Zhong Liu
DOI: https://doi.org/10.1055/s-0040-1710382
IF: 1.8
2020-01-01
Methods of Information in Medicine
Abstract:Background and Objectives The penetration rate of physical examinations in China is substantially lower than that in developed countries. Therefore, an auxiliary approach that does not depend on hospital health checks for the diagnosis of metabolic syndrome (MetS) is needed. Methods In this study, we proposed an augmented method with inferred blood features that uses self-care inputs available at home for the auxiliary diagnosis of MetS. The dataset used for modeling contained data on 91,420 individuals who had at least 2 consecutive years of health checks. We trained three separate models using a regularized gradient-boosted decision tree. The first model used only home-based features; additional blood test data (including triglyceride [TG] data, fasting blood glucose data, and high-density lipoprotein cholesterol [HDL-C] data) were included in the second model. However, in the augmented approach, the blood test data were manipulated using multivariate imputation by chained equations prior to inclusion in the third model. The performance of the three models for MetS auxiliary diagnosis was then quantitatively compared. Results The results showed that the third model exhibited the highest classification accuracy for MetS in comparison with the other two models (area under the curve [AUC]: 3rd vs. 2nd vs. 1st = 0.971 vs. 0.950 vs. 0.905,p < 0.001). We further revealed that with full sets of the three measurements from earlier blood test data, the classification accuracy of MetS can be further improved (AUC: without vs. with = 0.971 vs. 0.993). However, the magnitude of improvement was not statistically significant at the 1% level of significance (p = 0.014). Conclusion Our findings demonstrate the feasibility of the third model for MetS homecare applications and lend novel insights into innovative research on the health management of MetS. Further validation and implementation of our proposed model might improve quality of life and ultimately benefit the general population.